Huawei Cloud has launched its Pangu Models 5.5 at the Huawei Developer Conference 2025, aiming to address real-world business challenges through artificial intelligence. The latest version features enhancements in five essential areas: natural language processing (NLP), computer vision (CV), multimodal modeling, predictive analytics, and scientific computing. Zhang Ping’an, executive director of Huawei and CEO of Huawei Cloud, emphasized that the Pangu lineup is designed to provide AI solutions targeted at specific industries rather than generic tools.
A notable improvement in the NLP model is its 718-billion-parameter mixture of experts design. This model features 256 experts that can be deployed according to the specific task at hand. According to Huawei, this allows the model to excel in reasoning, tool usage, and mathematical problem-solving, handling longer text sequences while minimizing hallucinations.
An innovative feature is the adaptive setup, which provides swift responses for easier inquiries while allowing for deeper, more thoughtful responses to complex questions. This efficiency boost can enhance productivity by as much as eight times. Offering tools for crafting custom models, Huawei Cloud provides six core elements tailored for enterprise use, including foundational and industry-specific models, curated datasets, data engineering tools, and benchmarking platforms.
A practical example of this is the Chinese Academy of Agricultural Sciences (CAAS), which utilized Huawei’s models to enhance agricultural research, gaining the ability to analyze genes, conduct experiments, and develop improved rice varieties that are more wind-resistant. Additionally, Huawei is introducing models specific to five sectors: healthcare, finance, government, industrial manufacturing, and automotive, set to go live in June. In the automotive sector, for instance, the Pangu World Model uses synthetic data generation to enable more effective simulation and testing without the costs associated with real-world trials.
Moreover, the Pangu Prediction Model employs a triplet transformer architecture to analyze diverse data types, aiding companies like Conch Cement in optimizing their raw material usage while enhancing product quality. Similarly, local meteorological departments have integrated Pangu technology to enhance short-term rainfall predictions, benefiting energy companies in forecasting renewable power outputs. In terms of computer vision advancements, Huawei’s new 30-billion-parameter CV model can process various input types and detect subtle issues in industrial environments that typically escape human observation.
Notably, CNPC, China’s national oil company, has successfully implemented this model in maintenance and operational processes, significantly improving fault detection rates and reducing manual effort. Overall, these advancements in AI represent Huawei’s commitment to providing tailored solutions that cater to the specific needs of various industries, promoting enhanced efficiency and productivity across the board.